@pauliusztin_: I researched how Cognee, Graphiti and agent-memory (by Neo4j) built their agent-memory solutions and compiled the whole…
Summary
A summary of research comparing agent-memory solutions from Cognee, Graphiti, and Neo4j, which all use knowledge graphs and vector search with LLM-based data extraction.
View Cached Full Text
Cached at: 07/15/26, 10:00 PM
I researched how Cognee, Graphiti and agent-memory (by Neo4j) built their agent-memory solutions and compiled the whole architecture into a 2800-word article.
Their solution has huge overlaps:
-
It’s based on knowledge graphs (KG) + vector search
-
the data model and ontology sit at the core - data extraction is done via LLMs (+ some other tricks to reduce costs)
Similar Articles
rohitg00/agentmemory
agentmemory is an open-source persistent memory layer for AI coding agents (Claude Code, Cursor, Gemini CLI, Codex CLI, etc.) that uses knowledge graphs, confidence scoring, and hybrid search to give agents long-term memory across sessions via MCP, hooks, or REST API. Built on the iii engine, it requires no external databases and exposes 51 MCP tools.
@_avichawla: Build human-like memory for your Agents (open-source)! Every agentic and RAG system struggles with real-time knowledge …
Graphiti is an open-source tool that builds human-like memory for AI agents using a continuously evolving, temporally-aware knowledge graph, achieving up to 18.5% higher accuracy and 90% lower latency compared to MemGPT.
@akshay_pachaar: Your agent remembers everything and understands nothing. Most agent memory systems optimize for recall. The harder prob…
A thread discussing the importance of schema discipline in agent memory, introducing Zep AI's open-source Graphiti library for building temporal knowledge graphs with constrained entity and relationship types.
@yibie: Recommend this article. The teams from SJTU and Tsinghua systematically evaluated 12 agent memory systems. It's not one of those "our model is better" papers but rather breaks down how to choose memory systems from a data management perspective—when to use RAG, when to use vector databases, when to use knowledge graphs. Long-term memory for agents...
This paper from SJTU and Tsinghua systematically evaluates 12 agent memory systems from a data management perspective, decomposing memory into four modules and providing guidelines on when to use RAG, vector databases, or knowledge graphs for long-term agent memory.
Building Agentic GraphRAG Systems: From knowledge graphs and ontologies to a unified memory as an MCP server for your AI agent.
The author argues that GraphRAG is fundamentally a data modeling problem rather than just a retrieval algorithm, proposing a five-component architecture using ontologies, knowledge graphs, and an MCP server for unified agent memory.